Engineering · Platform
Operate the fleet — performance, policy, and the pipes.
Your seat answers: are agents responsive, governed, and instrumented — and what's the one thing to fix next?
Bottom line
Instrumentation coverage 88% (7/8 reporting; 1 shadow). Fix next: Codex CLI — worst tail at 6,800 ms p95, $78.00, ungoverned — one prioritized target. 3 agents are ungoverned. Latency measures response speed, not task success — agent outcome/quality is not yet measured.
88%
instrumentation coverage
7 of 8 reporting · 1 shadow
6,800 ms
worst p95 latency
Codex CLI
3
ungoverned agents
no policy binding
6
MCP servers
tool surfaces in use
Fix next — one prioritized target
worst p95 × cost × governanceCodex CLI is your slowest agent at 6,800 ms p95, costs $78.00, and is ungoverned.
Highest tail latency in the fleet — wiring or budget here moves the worst number first. Open Performance →
Also high-latency × high-cost: Codex CLI, Claude Code — slow and expensive, watch together.
Your surfaces
Is it working? — task success, not just speed & cost.
LLM responsiveness — tail latency.
Every agent + runtime detail.
What an enforcing policy would block.
Trace a run end to end when something breaks.
Stream telemetry to your stack.
Derived: instrumentation coverage · fix-next (worst p95 × cost × governed) — crossing performance.agents with agents in app/data/c16-fleet.ts.